Re: [libata/ahci] 8a4aeec8d2d: +138.4% perf-stat.dTLB-store-misses, +37.2% perf-stat.dTLB-load-misses

From: Dan Williams
Date: Tue Apr 22 2014 - 13:12:05 EST


On Mon, Apr 21, 2014 at 12:29 AM, Jet Chen <jet.chen@xxxxxxxxx> wrote:
> HI Dan,
>
> we noticed the below changes on
>
> git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata for-next
> commit 8a4aeec8d2d6a3edeffbdfae451cdf05cbf0fefd ("libata/ahci: accommodate
> tag ordered controllers")

Hi, was this on simulated hardware or a real AHCI controller and disk?

It does appear this test noticed increased throughput:

203893 ~ 0% +3.7% 211474 ~ 0% TOTAL iostat.sda.wkB/s

I wonder if ap->last_tag can be moved to a hotter cacheline, but if
throughput goes up I can imagine it throws off the cpu statistics
quite a bit.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/